summaryrefslogtreecommitdiffstats
path: root/image/SurfaceCache.cpp
diff options
context:
space:
mode:
authorDaniel Baumann <daniel.baumann@progress-linux.org>2024-04-28 14:29:10 +0000
committerDaniel Baumann <daniel.baumann@progress-linux.org>2024-04-28 14:29:10 +0000
commit2aa4a82499d4becd2284cdb482213d541b8804dd (patch)
treeb80bf8bf13c3766139fbacc530efd0dd9d54394c /image/SurfaceCache.cpp
parentInitial commit. (diff)
downloadfirefox-2aa4a82499d4becd2284cdb482213d541b8804dd.tar.xz
firefox-2aa4a82499d4becd2284cdb482213d541b8804dd.zip
Adding upstream version 86.0.1.upstream/86.0.1upstream
Signed-off-by: Daniel Baumann <daniel.baumann@progress-linux.org>
Diffstat (limited to 'image/SurfaceCache.cpp')
-rw-r--r--image/SurfaceCache.cpp1846
1 files changed, 1846 insertions, 0 deletions
diff --git a/image/SurfaceCache.cpp b/image/SurfaceCache.cpp
new file mode 100644
index 0000000000..2d058b41be
--- /dev/null
+++ b/image/SurfaceCache.cpp
@@ -0,0 +1,1846 @@
+/* -*- Mode: C++; tab-width: 2; indent-tabs-mode: nil; c-basic-offset: 2 -*- */
+/* This Source Code Form is subject to the terms of the Mozilla Public
+ * License, v. 2.0. If a copy of the MPL was not distributed with this
+ * file, You can obtain one at http://mozilla.org/MPL/2.0/. */
+
+/**
+ * SurfaceCache is a service for caching temporary surfaces in imagelib.
+ */
+
+#include "SurfaceCache.h"
+
+#include <algorithm>
+#include <utility>
+
+#include "ISurfaceProvider.h"
+#include "Image.h"
+#include "LookupResult.h"
+#include "ShutdownTracker.h"
+#include "gfx2DGlue.h"
+#include "gfxPlatform.h"
+#include "imgFrame.h"
+#include "mozilla/Assertions.h"
+#include "mozilla/Attributes.h"
+#include "mozilla/CheckedInt.h"
+#include "mozilla/DebugOnly.h"
+#include "mozilla/Likely.h"
+#include "mozilla/RefPtr.h"
+#include "mozilla/StaticMutex.h"
+#include "mozilla/StaticPrefs_image.h"
+#include "mozilla/StaticPtr.h"
+#include "mozilla/Tuple.h"
+#include "nsExpirationTracker.h"
+#include "nsHashKeys.h"
+#include "nsIMemoryReporter.h"
+#include "nsRefPtrHashtable.h"
+#include "nsSize.h"
+#include "nsTArray.h"
+#include "Orientation.h"
+#include "prsystem.h"
+
+using std::max;
+using std::min;
+
+namespace mozilla {
+
+using namespace gfx;
+
+namespace image {
+
+MOZ_DEFINE_MALLOC_SIZE_OF(SurfaceCacheMallocSizeOf)
+
+class CachedSurface;
+class SurfaceCacheImpl;
+
+///////////////////////////////////////////////////////////////////////////////
+// Static Data
+///////////////////////////////////////////////////////////////////////////////
+
+// The single surface cache instance.
+static StaticRefPtr<SurfaceCacheImpl> sInstance;
+
+// The mutex protecting the surface cache.
+static StaticMutex sInstanceMutex;
+
+///////////////////////////////////////////////////////////////////////////////
+// SurfaceCache Implementation
+///////////////////////////////////////////////////////////////////////////////
+
+/**
+ * Cost models the cost of storing a surface in the cache. Right now, this is
+ * simply an estimate of the size of the surface in bytes, but in the future it
+ * may be worth taking into account the cost of rematerializing the surface as
+ * well.
+ */
+typedef size_t Cost;
+
+static Cost ComputeCost(const IntSize& aSize, uint32_t aBytesPerPixel) {
+ MOZ_ASSERT(aBytesPerPixel == 1 || aBytesPerPixel == 4);
+ return aSize.width * aSize.height * aBytesPerPixel;
+}
+
+/**
+ * Since we want to be able to make eviction decisions based on cost, we need to
+ * be able to look up the CachedSurface which has a certain cost as well as the
+ * cost associated with a certain CachedSurface. To make this possible, in data
+ * structures we actually store a CostEntry, which contains a weak pointer to
+ * its associated surface.
+ *
+ * To make usage of the weak pointer safe, SurfaceCacheImpl always calls
+ * StartTracking after a surface is stored in the cache and StopTracking before
+ * it is removed.
+ */
+class CostEntry {
+ public:
+ CostEntry(NotNull<CachedSurface*> aSurface, Cost aCost)
+ : mSurface(aSurface), mCost(aCost) {}
+
+ NotNull<CachedSurface*> Surface() const { return mSurface; }
+ Cost GetCost() const { return mCost; }
+
+ bool operator==(const CostEntry& aOther) const {
+ return mSurface == aOther.mSurface && mCost == aOther.mCost;
+ }
+
+ bool operator<(const CostEntry& aOther) const {
+ return mCost < aOther.mCost ||
+ (mCost == aOther.mCost && mSurface < aOther.mSurface);
+ }
+
+ private:
+ NotNull<CachedSurface*> mSurface;
+ Cost mCost;
+};
+
+/**
+ * A CachedSurface associates a surface with a key that uniquely identifies that
+ * surface.
+ */
+class CachedSurface {
+ ~CachedSurface() {}
+
+ public:
+ MOZ_DECLARE_REFCOUNTED_TYPENAME(CachedSurface)
+ NS_INLINE_DECL_THREADSAFE_REFCOUNTING(CachedSurface)
+
+ explicit CachedSurface(NotNull<ISurfaceProvider*> aProvider)
+ : mProvider(aProvider), mIsLocked(false) {}
+
+ DrawableSurface GetDrawableSurface() const {
+ if (MOZ_UNLIKELY(IsPlaceholder())) {
+ MOZ_ASSERT_UNREACHABLE("Called GetDrawableSurface() on a placeholder");
+ return DrawableSurface();
+ }
+
+ return mProvider->Surface();
+ }
+
+ void SetLocked(bool aLocked) {
+ if (IsPlaceholder()) {
+ return; // Can't lock a placeholder.
+ }
+
+ // Update both our state and our provider's state. Some surface providers
+ // are permanently locked; maintaining our own locking state enables us to
+ // respect SetLocked() even when it's meaningless from the provider's
+ // perspective.
+ mIsLocked = aLocked;
+ mProvider->SetLocked(aLocked);
+ }
+
+ bool IsLocked() const {
+ return !IsPlaceholder() && mIsLocked && mProvider->IsLocked();
+ }
+
+ void SetCannotSubstitute() {
+ mProvider->Availability().SetCannotSubstitute();
+ }
+ bool CannotSubstitute() const {
+ return mProvider->Availability().CannotSubstitute();
+ }
+
+ bool IsPlaceholder() const {
+ return mProvider->Availability().IsPlaceholder();
+ }
+ bool IsDecoded() const { return !IsPlaceholder() && mProvider->IsFinished(); }
+
+ ImageKey GetImageKey() const { return mProvider->GetImageKey(); }
+ const SurfaceKey& GetSurfaceKey() const { return mProvider->GetSurfaceKey(); }
+ nsExpirationState* GetExpirationState() { return &mExpirationState; }
+
+ CostEntry GetCostEntry() {
+ return image::CostEntry(WrapNotNull(this), mProvider->LogicalSizeInBytes());
+ }
+
+ size_t ShallowSizeOfIncludingThis(MallocSizeOf aMallocSizeOf) const {
+ return aMallocSizeOf(this) + aMallocSizeOf(mProvider.get());
+ }
+
+ // A helper type used by SurfaceCacheImpl::CollectSizeOfSurfaces.
+ struct MOZ_STACK_CLASS SurfaceMemoryReport {
+ SurfaceMemoryReport(nsTArray<SurfaceMemoryCounter>& aCounters,
+ MallocSizeOf aMallocSizeOf)
+ : mCounters(aCounters), mMallocSizeOf(aMallocSizeOf) {}
+
+ void Add(NotNull<CachedSurface*> aCachedSurface, bool aIsFactor2) {
+ if (aCachedSurface->IsPlaceholder()) {
+ return;
+ }
+
+ // Record the memory used by the ISurfaceProvider. This may not have a
+ // straightforward relationship to the size of the surface that
+ // DrawableRef() returns if the surface is generated dynamically. (i.e.,
+ // for surfaces with PlaybackType::eAnimated.)
+ aCachedSurface->mProvider->AddSizeOfExcludingThis(
+ mMallocSizeOf, [&](ISurfaceProvider::AddSizeOfCbData& aMetadata) {
+ SurfaceMemoryCounter counter(aCachedSurface->GetSurfaceKey(),
+ aCachedSurface->IsLocked(),
+ aCachedSurface->CannotSubstitute(),
+ aIsFactor2, aMetadata.mFinished);
+
+ counter.Values().SetDecodedHeap(aMetadata.mHeapBytes);
+ counter.Values().SetDecodedNonHeap(aMetadata.mNonHeapBytes);
+ counter.Values().SetDecodedUnknown(aMetadata.mUnknownBytes);
+ counter.Values().SetExternalHandles(aMetadata.mExternalHandles);
+ counter.Values().SetFrameIndex(aMetadata.mIndex);
+ counter.Values().SetExternalId(aMetadata.mExternalId);
+ counter.Values().SetSurfaceTypes(aMetadata.mTypes);
+
+ mCounters.AppendElement(counter);
+ });
+ }
+
+ private:
+ nsTArray<SurfaceMemoryCounter>& mCounters;
+ MallocSizeOf mMallocSizeOf;
+ };
+
+ private:
+ nsExpirationState mExpirationState;
+ NotNull<RefPtr<ISurfaceProvider>> mProvider;
+ bool mIsLocked;
+};
+
+static int64_t AreaOfIntSize(const IntSize& aSize) {
+ return static_cast<int64_t>(aSize.width) * static_cast<int64_t>(aSize.height);
+}
+
+/**
+ * An ImageSurfaceCache is a per-image surface cache. For correctness we must be
+ * able to remove all surfaces associated with an image when the image is
+ * destroyed or invalidated. Since this will happen frequently, it makes sense
+ * to make it cheap by storing the surfaces for each image separately.
+ *
+ * ImageSurfaceCache also keeps track of whether its associated image is locked
+ * or unlocked.
+ *
+ * The cache may also enter "factor of 2" mode which occurs when the number of
+ * surfaces in the cache exceeds the "image.cache.factor2.threshold-surfaces"
+ * pref plus the number of native sizes of the image. When in "factor of 2"
+ * mode, the cache will strongly favour sizes which are a factor of 2 of the
+ * largest native size. It accomplishes this by suggesting a factor of 2 size
+ * when lookups fail and substituting the nearest factor of 2 surface to the
+ * ideal size as the "best" available (as opposed to substitution but not
+ * found). This allows us to minimize memory consumption and CPU time spent
+ * decoding when a website requires many variants of the same surface.
+ */
+class ImageSurfaceCache {
+ ~ImageSurfaceCache() {}
+
+ public:
+ explicit ImageSurfaceCache(const ImageKey aImageKey)
+ : mLocked(false),
+ mFactor2Mode(false),
+ mFactor2Pruned(false),
+ mIsVectorImage(aImageKey->GetType() == imgIContainer::TYPE_VECTOR) {}
+
+ MOZ_DECLARE_REFCOUNTED_TYPENAME(ImageSurfaceCache)
+ NS_INLINE_DECL_THREADSAFE_REFCOUNTING(ImageSurfaceCache)
+
+ typedef nsRefPtrHashtable<nsGenericHashKey<SurfaceKey>, CachedSurface>
+ SurfaceTable;
+
+ bool IsEmpty() const { return mSurfaces.Count() == 0; }
+
+ size_t ShallowSizeOfIncludingThis(MallocSizeOf aMallocSizeOf) const {
+ size_t bytes = aMallocSizeOf(this) +
+ mSurfaces.ShallowSizeOfExcludingThis(aMallocSizeOf);
+ for (auto iter = ConstIter(); !iter.Done(); iter.Next()) {
+ bytes += iter.UserData()->ShallowSizeOfIncludingThis(aMallocSizeOf);
+ }
+ return bytes;
+ }
+
+ [[nodiscard]] bool Insert(NotNull<CachedSurface*> aSurface) {
+ MOZ_ASSERT(!mLocked || aSurface->IsPlaceholder() || aSurface->IsLocked(),
+ "Inserting an unlocked surface for a locked image");
+ return mSurfaces.Put(aSurface->GetSurfaceKey(),
+ RefPtr<CachedSurface>{aSurface}, fallible);
+ }
+
+ already_AddRefed<CachedSurface> Remove(NotNull<CachedSurface*> aSurface) {
+ MOZ_ASSERT(mSurfaces.GetWeak(aSurface->GetSurfaceKey()),
+ "Should not be removing a surface we don't have");
+
+ RefPtr<CachedSurface> surface;
+ mSurfaces.Remove(aSurface->GetSurfaceKey(), getter_AddRefs(surface));
+ AfterMaybeRemove();
+ return surface.forget();
+ }
+
+ already_AddRefed<CachedSurface> Lookup(const SurfaceKey& aSurfaceKey,
+ bool aForAccess) {
+ RefPtr<CachedSurface> surface;
+ mSurfaces.Get(aSurfaceKey, getter_AddRefs(surface));
+
+ if (aForAccess) {
+ if (surface) {
+ // We don't want to allow factor of 2 mode pruning to release surfaces
+ // for which the callers will accept no substitute.
+ surface->SetCannotSubstitute();
+ } else if (!mFactor2Mode) {
+ // If no exact match is found, and this is for use rather than internal
+ // accounting (i.e. insert and removal), we know this will trigger a
+ // decode. Make sure we switch now to factor of 2 mode if necessary.
+ MaybeSetFactor2Mode();
+ }
+ }
+
+ return surface.forget();
+ }
+
+ /**
+ * @returns A tuple containing the best matching CachedSurface if available,
+ * a MatchType describing how the CachedSurface was selected, and
+ * an IntSize which is the size the caller should choose to decode
+ * at should it attempt to do so.
+ */
+ Tuple<already_AddRefed<CachedSurface>, MatchType, IntSize> LookupBestMatch(
+ const SurfaceKey& aIdealKey) {
+ // Try for an exact match first.
+ RefPtr<CachedSurface> exactMatch;
+ mSurfaces.Get(aIdealKey, getter_AddRefs(exactMatch));
+ if (exactMatch) {
+ if (exactMatch->IsDecoded()) {
+ return MakeTuple(exactMatch.forget(), MatchType::EXACT, IntSize());
+ }
+ } else if (!mFactor2Mode) {
+ // If no exact match is found, and we are not in factor of 2 mode, then
+ // we know that we will trigger a decode because at best we will provide
+ // a substitute. Make sure we switch now to factor of 2 mode if necessary.
+ MaybeSetFactor2Mode();
+ }
+
+ // Try for a best match second, if using compact.
+ IntSize suggestedSize = SuggestedSize(aIdealKey.Size());
+ if (suggestedSize != aIdealKey.Size()) {
+ if (!exactMatch) {
+ SurfaceKey compactKey = aIdealKey.CloneWithSize(suggestedSize);
+ mSurfaces.Get(compactKey, getter_AddRefs(exactMatch));
+ if (exactMatch && exactMatch->IsDecoded()) {
+ MOZ_ASSERT(suggestedSize != aIdealKey.Size());
+ return MakeTuple(exactMatch.forget(),
+ MatchType::SUBSTITUTE_BECAUSE_BEST, suggestedSize);
+ }
+ }
+ }
+
+ // There's no perfect match, so find the best match we can.
+ RefPtr<CachedSurface> bestMatch;
+ for (auto iter = ConstIter(); !iter.Done(); iter.Next()) {
+ NotNull<CachedSurface*> current = WrapNotNull(iter.UserData());
+ const SurfaceKey& currentKey = current->GetSurfaceKey();
+
+ // We never match a placeholder.
+ if (current->IsPlaceholder()) {
+ continue;
+ }
+ // Matching the playback type and SVG context is required.
+ if (currentKey.Playback() != aIdealKey.Playback() ||
+ currentKey.SVGContext() != aIdealKey.SVGContext()) {
+ continue;
+ }
+ // Matching the flags is required.
+ if (currentKey.Flags() != aIdealKey.Flags()) {
+ continue;
+ }
+ // Anything is better than nothing! (Within the constraints we just
+ // checked, of course.)
+ if (!bestMatch) {
+ bestMatch = current;
+ continue;
+ }
+
+ MOZ_ASSERT(bestMatch, "Should have a current best match");
+
+ // Always prefer completely decoded surfaces.
+ bool bestMatchIsDecoded = bestMatch->IsDecoded();
+ if (bestMatchIsDecoded && !current->IsDecoded()) {
+ continue;
+ }
+ if (!bestMatchIsDecoded && current->IsDecoded()) {
+ bestMatch = current;
+ continue;
+ }
+
+ SurfaceKey bestMatchKey = bestMatch->GetSurfaceKey();
+ if (CompareArea(aIdealKey.Size(), bestMatchKey.Size(),
+ currentKey.Size())) {
+ bestMatch = current;
+ }
+ }
+
+ MatchType matchType;
+ if (bestMatch) {
+ if (!exactMatch) {
+ // No exact match, neither ideal nor factor of 2.
+ MOZ_ASSERT(suggestedSize != bestMatch->GetSurfaceKey().Size(),
+ "No exact match despite the fact the sizes match!");
+ matchType = MatchType::SUBSTITUTE_BECAUSE_NOT_FOUND;
+ } else if (exactMatch != bestMatch) {
+ // The exact match is still decoding, but we found a substitute.
+ matchType = MatchType::SUBSTITUTE_BECAUSE_PENDING;
+ } else if (aIdealKey.Size() != bestMatch->GetSurfaceKey().Size()) {
+ // The best factor of 2 match is still decoding, but the best we've got.
+ MOZ_ASSERT(suggestedSize != aIdealKey.Size());
+ MOZ_ASSERT(mFactor2Mode || mIsVectorImage);
+ matchType = MatchType::SUBSTITUTE_BECAUSE_BEST;
+ } else {
+ // The exact match is still decoding, but it's the best we've got.
+ matchType = MatchType::EXACT;
+ }
+ } else {
+ if (exactMatch) {
+ // We found an "exact match"; it must have been a placeholder.
+ MOZ_ASSERT(exactMatch->IsPlaceholder());
+ matchType = MatchType::PENDING;
+ } else {
+ // We couldn't find an exact match *or* a substitute.
+ matchType = MatchType::NOT_FOUND;
+ }
+ }
+
+ return MakeTuple(bestMatch.forget(), matchType, suggestedSize);
+ }
+
+ void MaybeSetFactor2Mode() {
+ MOZ_ASSERT(!mFactor2Mode);
+
+ // Typically an image cache will not have too many size-varying surfaces, so
+ // if we exceed the given threshold, we should consider using a subset.
+ int32_t thresholdSurfaces =
+ StaticPrefs::image_cache_factor2_threshold_surfaces();
+ if (thresholdSurfaces < 0 ||
+ mSurfaces.Count() <= static_cast<uint32_t>(thresholdSurfaces)) {
+ return;
+ }
+
+ // Determine how many native surfaces this image has. If it is zero, and it
+ // is a vector image, then we should impute a single native size. Otherwise,
+ // it may be zero because we don't know yet, or the image has an error, or
+ // it isn't supported.
+ auto first = ConstIter();
+ NotNull<CachedSurface*> current = WrapNotNull(first.UserData());
+ Image* image = static_cast<Image*>(current->GetImageKey());
+ size_t nativeSizes = image->GetNativeSizesLength();
+ if (mIsVectorImage) {
+ MOZ_ASSERT(nativeSizes == 0);
+ nativeSizes = 1;
+ } else if (nativeSizes == 0) {
+ return;
+ }
+
+ // Increase the threshold by the number of native sizes. This ensures that
+ // we do not prevent decoding of the image at all its native sizes. It does
+ // not guarantee we will provide a surface at that size however (i.e. many
+ // other sized surfaces are requested, in addition to the native sizes).
+ thresholdSurfaces += nativeSizes;
+ if (mSurfaces.Count() <= static_cast<uint32_t>(thresholdSurfaces)) {
+ return;
+ }
+
+ // Get our native size. While we know the image should be fully decoded,
+ // if it is an SVG, it is valid to have a zero size. We can't do compacting
+ // in that case because we need to know the width/height ratio to define a
+ // candidate set.
+ IntSize nativeSize;
+ if (NS_FAILED(image->GetWidth(&nativeSize.width)) ||
+ NS_FAILED(image->GetHeight(&nativeSize.height)) ||
+ nativeSize.IsEmpty()) {
+ return;
+ }
+
+ // We have a valid size, we can change modes.
+ mFactor2Mode = true;
+ }
+
+ template <typename Function>
+ void Prune(Function&& aRemoveCallback) {
+ if (!mFactor2Mode || mFactor2Pruned) {
+ return;
+ }
+
+ // Attempt to discard any surfaces which are not factor of 2 and the best
+ // factor of 2 match exists.
+ bool hasNotFactorSize = false;
+ for (auto iter = mSurfaces.Iter(); !iter.Done(); iter.Next()) {
+ NotNull<CachedSurface*> current = WrapNotNull(iter.UserData());
+ const SurfaceKey& currentKey = current->GetSurfaceKey();
+ const IntSize& currentSize = currentKey.Size();
+
+ // First we check if someone requested this size and would not accept
+ // an alternatively sized surface.
+ if (current->CannotSubstitute()) {
+ continue;
+ }
+
+ // Next we find the best factor of 2 size for this surface. If this
+ // surface is a factor of 2 size, then we want to keep it.
+ IntSize bestSize = SuggestedSize(currentSize);
+ if (bestSize == currentSize) {
+ continue;
+ }
+
+ // Check the cache for a surface with the same parameters except for the
+ // size which uses the closest factor of 2 size.
+ SurfaceKey compactKey = currentKey.CloneWithSize(bestSize);
+ RefPtr<CachedSurface> compactMatch;
+ mSurfaces.Get(compactKey, getter_AddRefs(compactMatch));
+ if (compactMatch && compactMatch->IsDecoded()) {
+ aRemoveCallback(current);
+ iter.Remove();
+ } else {
+ hasNotFactorSize = true;
+ }
+ }
+
+ // We have no surfaces that are not factor of 2 sized, so we can stop
+ // pruning henceforth, because we avoid the insertion of new surfaces that
+ // don't match our sizing set (unless the caller won't accept a
+ // substitution.)
+ if (!hasNotFactorSize) {
+ mFactor2Pruned = true;
+ }
+
+ // We should never leave factor of 2 mode due to pruning in of itself, but
+ // if we discarded surfaces due to the volatile buffers getting released,
+ // it is possible.
+ AfterMaybeRemove();
+ }
+
+ IntSize SuggestedSize(const IntSize& aSize) const {
+ IntSize suggestedSize = SuggestedSizeInternal(aSize);
+ if (mIsVectorImage) {
+ suggestedSize = SurfaceCache::ClampVectorSize(suggestedSize);
+ }
+ return suggestedSize;
+ }
+
+ IntSize SuggestedSizeInternal(const IntSize& aSize) const {
+ // When not in factor of 2 mode, we can always decode at the given size.
+ if (!mFactor2Mode) {
+ return aSize;
+ }
+
+ // We cannot enter factor of 2 mode unless we have a minimum number of
+ // surfaces, and we should have left it if the cache was emptied.
+ if (MOZ_UNLIKELY(IsEmpty())) {
+ MOZ_ASSERT_UNREACHABLE("Should not be empty and in factor of 2 mode!");
+ return aSize;
+ }
+
+ // This bit of awkwardness gets the largest native size of the image.
+ auto iter = ConstIter();
+ NotNull<CachedSurface*> firstSurface = WrapNotNull(iter.UserData());
+ Image* image = static_cast<Image*>(firstSurface->GetImageKey());
+ IntSize factorSize;
+ if (NS_FAILED(image->GetWidth(&factorSize.width)) ||
+ NS_FAILED(image->GetHeight(&factorSize.height)) ||
+ factorSize.IsEmpty()) {
+ // We should not have entered factor of 2 mode without a valid size, and
+ // several successfully decoded surfaces. Note that valid vector images
+ // may have a default size of 0x0, and those are not yet supported.
+ MOZ_ASSERT_UNREACHABLE("Expected valid native size!");
+ return aSize;
+ }
+ if (image->GetOrientation().SwapsWidthAndHeight() &&
+ image->HandledOrientation()) {
+ std::swap(factorSize.width, factorSize.height);
+ }
+
+ if (mIsVectorImage) {
+ // Ensure the aspect ratio matches the native size before forcing the
+ // caller to accept a factor of 2 size. The difference between the aspect
+ // ratios is:
+ //
+ // delta = nativeWidth/nativeHeight - desiredWidth/desiredHeight
+ //
+ // delta*nativeHeight*desiredHeight = nativeWidth*desiredHeight
+ // - desiredWidth*nativeHeight
+ //
+ // Using the maximum accepted delta as a constant, we can avoid the
+ // floating point division and just compare after some integer ops.
+ int32_t delta =
+ factorSize.width * aSize.height - aSize.width * factorSize.height;
+ int32_t maxDelta = (factorSize.height * aSize.height) >> 4;
+ if (delta > maxDelta || delta < -maxDelta) {
+ return aSize;
+ }
+
+ // If the requested size is bigger than the native size, we actually need
+ // to grow the native size instead of shrinking it.
+ if (factorSize.width < aSize.width) {
+ do {
+ IntSize candidate(factorSize.width * 2, factorSize.height * 2);
+ if (!SurfaceCache::IsLegalSize(candidate)) {
+ break;
+ }
+
+ factorSize = candidate;
+ } while (factorSize.width < aSize.width);
+
+ return factorSize;
+ }
+
+ // Otherwise we can find the best fit as normal.
+ }
+
+ // Start with the native size as the best first guess.
+ IntSize bestSize = factorSize;
+ factorSize.width /= 2;
+ factorSize.height /= 2;
+
+ while (!factorSize.IsEmpty()) {
+ if (!CompareArea(aSize, bestSize, factorSize)) {
+ // This size is not better than the last. Since we proceed from largest
+ // to smallest, we know that the next size will not be better if the
+ // previous size was rejected. Break early.
+ break;
+ }
+
+ // The current factor of 2 size is better than the last selected size.
+ bestSize = factorSize;
+ factorSize.width /= 2;
+ factorSize.height /= 2;
+ }
+
+ return bestSize;
+ }
+
+ bool CompareArea(const IntSize& aIdealSize, const IntSize& aBestSize,
+ const IntSize& aSize) const {
+ // Compare sizes. We use an area-based heuristic here instead of computing a
+ // truly optimal answer, since it seems very unlikely to make a difference
+ // for realistic sizes.
+ int64_t idealArea = AreaOfIntSize(aIdealSize);
+ int64_t currentArea = AreaOfIntSize(aSize);
+ int64_t bestMatchArea = AreaOfIntSize(aBestSize);
+
+ // If the best match is smaller than the ideal size, prefer bigger sizes.
+ if (bestMatchArea < idealArea) {
+ if (currentArea > bestMatchArea) {
+ return true;
+ }
+ return false;
+ }
+
+ // Other, prefer sizes closer to the ideal size, but still not smaller.
+ if (idealArea <= currentArea && currentArea < bestMatchArea) {
+ return true;
+ }
+
+ // This surface isn't an improvement over the current best match.
+ return false;
+ }
+
+ template <typename Function>
+ void CollectSizeOfSurfaces(nsTArray<SurfaceMemoryCounter>& aCounters,
+ MallocSizeOf aMallocSizeOf,
+ Function&& aRemoveCallback) {
+ CachedSurface::SurfaceMemoryReport report(aCounters, aMallocSizeOf);
+ for (auto iter = mSurfaces.Iter(); !iter.Done(); iter.Next()) {
+ NotNull<CachedSurface*> surface = WrapNotNull(iter.UserData());
+
+ // We don't need the drawable surface for ourselves, but adding a surface
+ // to the report will trigger this indirectly. If the surface was
+ // discarded by the OS because it was in volatile memory, we should remove
+ // it from the cache immediately rather than include it in the report.
+ DrawableSurface drawableSurface;
+ if (!surface->IsPlaceholder()) {
+ drawableSurface = surface->GetDrawableSurface();
+ if (!drawableSurface) {
+ aRemoveCallback(surface);
+ iter.Remove();
+ continue;
+ }
+ }
+
+ const IntSize& size = surface->GetSurfaceKey().Size();
+ bool factor2Size = false;
+ if (mFactor2Mode) {
+ factor2Size = (size == SuggestedSize(size));
+ }
+ report.Add(surface, factor2Size);
+ }
+
+ AfterMaybeRemove();
+ }
+
+ SurfaceTable::Iterator ConstIter() const { return mSurfaces.ConstIter(); }
+ uint32_t Count() const { return mSurfaces.Count(); }
+
+ void SetLocked(bool aLocked) { mLocked = aLocked; }
+ bool IsLocked() const { return mLocked; }
+
+ private:
+ void AfterMaybeRemove() {
+ if (IsEmpty() && mFactor2Mode) {
+ // The last surface for this cache was removed. This can happen if the
+ // surface was stored in a volatile buffer and got purged, or the surface
+ // expired from the cache. If the cache itself lingers for some reason
+ // (e.g. in the process of performing a lookup, the cache itself is
+ // locked), then we need to reset the factor of 2 state because it
+ // requires at least one surface present to get the native size
+ // information from the image.
+ mFactor2Mode = mFactor2Pruned = false;
+ }
+ }
+
+ SurfaceTable mSurfaces;
+
+ bool mLocked;
+
+ // True in "factor of 2" mode.
+ bool mFactor2Mode;
+
+ // True if all non-factor of 2 surfaces have been removed from the cache. Note
+ // that this excludes unsubstitutable sizes.
+ bool mFactor2Pruned;
+
+ // True if the surfaces are produced from a vector image. If so, it must match
+ // the aspect ratio when using factor of 2 mode.
+ bool mIsVectorImage;
+};
+
+/**
+ * SurfaceCacheImpl is responsible for determining which surfaces will be cached
+ * and managing the surface cache data structures. Rather than interact with
+ * SurfaceCacheImpl directly, client code interacts with SurfaceCache, which
+ * maintains high-level invariants and encapsulates the details of the surface
+ * cache's implementation.
+ */
+class SurfaceCacheImpl final : public nsIMemoryReporter {
+ public:
+ NS_DECL_ISUPPORTS
+
+ SurfaceCacheImpl(uint32_t aSurfaceCacheExpirationTimeMS,
+ uint32_t aSurfaceCacheDiscardFactor,
+ uint32_t aSurfaceCacheSize)
+ : mExpirationTracker(aSurfaceCacheExpirationTimeMS),
+ mMemoryPressureObserver(new MemoryPressureObserver),
+ mDiscardFactor(aSurfaceCacheDiscardFactor),
+ mMaxCost(aSurfaceCacheSize),
+ mAvailableCost(aSurfaceCacheSize),
+ mLockedCost(0),
+ mOverflowCount(0),
+ mAlreadyPresentCount(0),
+ mTableFailureCount(0),
+ mTrackingFailureCount(0) {
+ nsCOMPtr<nsIObserverService> os = services::GetObserverService();
+ if (os) {
+ os->AddObserver(mMemoryPressureObserver, "memory-pressure", false);
+ }
+ }
+
+ private:
+ virtual ~SurfaceCacheImpl() {
+ nsCOMPtr<nsIObserverService> os = services::GetObserverService();
+ if (os) {
+ os->RemoveObserver(mMemoryPressureObserver, "memory-pressure");
+ }
+
+ UnregisterWeakMemoryReporter(this);
+ }
+
+ public:
+ void InitMemoryReporter() { RegisterWeakMemoryReporter(this); }
+
+ InsertOutcome Insert(NotNull<ISurfaceProvider*> aProvider, bool aSetAvailable,
+ const StaticMutexAutoLock& aAutoLock) {
+ // If this is a duplicate surface, refuse to replace the original.
+ // XXX(seth): Calling Lookup() and then RemoveEntry() does the lookup
+ // twice. We'll make this more efficient in bug 1185137.
+ LookupResult result =
+ Lookup(aProvider->GetImageKey(), aProvider->GetSurfaceKey(), aAutoLock,
+ /* aMarkUsed = */ false);
+ if (MOZ_UNLIKELY(result)) {
+ mAlreadyPresentCount++;
+ return InsertOutcome::FAILURE_ALREADY_PRESENT;
+ }
+
+ if (result.Type() == MatchType::PENDING) {
+ RemoveEntry(aProvider->GetImageKey(), aProvider->GetSurfaceKey(),
+ aAutoLock);
+ }
+
+ MOZ_ASSERT(result.Type() == MatchType::NOT_FOUND ||
+ result.Type() == MatchType::PENDING,
+ "A LookupResult with no surface should be NOT_FOUND or PENDING");
+
+ // If this is bigger than we can hold after discarding everything we can,
+ // refuse to cache it.
+ Cost cost = aProvider->LogicalSizeInBytes();
+ if (MOZ_UNLIKELY(!CanHoldAfterDiscarding(cost))) {
+ mOverflowCount++;
+ return InsertOutcome::FAILURE;
+ }
+
+ // Remove elements in order of cost until we can fit this in the cache. Note
+ // that locked surfaces aren't in mCosts, so we never remove them here.
+ while (cost > mAvailableCost) {
+ MOZ_ASSERT(!mCosts.IsEmpty(),
+ "Removed everything and it still won't fit");
+ Remove(mCosts.LastElement().Surface(), /* aStopTracking */ true,
+ aAutoLock);
+ }
+
+ // Locate the appropriate per-image cache. If there's not an existing cache
+ // for this image, create it.
+ const ImageKey imageKey = aProvider->GetImageKey();
+ RefPtr<ImageSurfaceCache> cache = GetImageCache(imageKey);
+ if (!cache) {
+ cache = new ImageSurfaceCache(imageKey);
+ if (!mImageCaches.Put(aProvider->GetImageKey(), RefPtr{cache},
+ fallible)) {
+ mTableFailureCount++;
+ return InsertOutcome::FAILURE;
+ }
+ }
+
+ // If we were asked to mark the cache entry available, do so.
+ if (aSetAvailable) {
+ aProvider->Availability().SetAvailable();
+ }
+
+ auto surface = MakeNotNull<RefPtr<CachedSurface>>(aProvider);
+
+ // We require that locking succeed if the image is locked and we're not
+ // inserting a placeholder; the caller may need to know this to handle
+ // errors correctly.
+ bool mustLock = cache->IsLocked() && !surface->IsPlaceholder();
+ if (mustLock) {
+ surface->SetLocked(true);
+ if (!surface->IsLocked()) {
+ return InsertOutcome::FAILURE;
+ }
+ }
+
+ // Insert.
+ MOZ_ASSERT(cost <= mAvailableCost, "Inserting despite too large a cost");
+ if (!cache->Insert(surface)) {
+ mTableFailureCount++;
+ if (mustLock) {
+ surface->SetLocked(false);
+ }
+ return InsertOutcome::FAILURE;
+ }
+
+ if (MOZ_UNLIKELY(!StartTracking(surface, aAutoLock))) {
+ MOZ_ASSERT(!mustLock);
+ Remove(surface, /* aStopTracking */ false, aAutoLock);
+ return InsertOutcome::FAILURE;
+ }
+
+ return InsertOutcome::SUCCESS;
+ }
+
+ void Remove(NotNull<CachedSurface*> aSurface, bool aStopTracking,
+ const StaticMutexAutoLock& aAutoLock) {
+ ImageKey imageKey = aSurface->GetImageKey();
+
+ RefPtr<ImageSurfaceCache> cache = GetImageCache(imageKey);
+ MOZ_ASSERT(cache, "Shouldn't try to remove a surface with no image cache");
+
+ // If the surface was not a placeholder, tell its image that we discarded
+ // it.
+ if (!aSurface->IsPlaceholder()) {
+ static_cast<Image*>(imageKey)->OnSurfaceDiscarded(
+ aSurface->GetSurfaceKey());
+ }
+
+ // If we failed during StartTracking, we can skip this step.
+ if (aStopTracking) {
+ StopTracking(aSurface, /* aIsTracked */ true, aAutoLock);
+ }
+
+ // Individual surfaces must be freed outside the lock.
+ mCachedSurfacesDiscard.AppendElement(cache->Remove(aSurface));
+
+ MaybeRemoveEmptyCache(imageKey, cache);
+ }
+
+ bool StartTracking(NotNull<CachedSurface*> aSurface,
+ const StaticMutexAutoLock& aAutoLock) {
+ CostEntry costEntry = aSurface->GetCostEntry();
+ MOZ_ASSERT(costEntry.GetCost() <= mAvailableCost,
+ "Cost too large and the caller didn't catch it");
+
+ if (aSurface->IsLocked()) {
+ mLockedCost += costEntry.GetCost();
+ MOZ_ASSERT(mLockedCost <= mMaxCost, "Locked more than we can hold?");
+ } else {
+ if (NS_WARN_IF(!mCosts.InsertElementSorted(costEntry, fallible))) {
+ mTrackingFailureCount++;
+ return false;
+ }
+
+ // This may fail during XPCOM shutdown, so we need to ensure the object is
+ // tracked before calling RemoveObject in StopTracking.
+ nsresult rv = mExpirationTracker.AddObjectLocked(aSurface, aAutoLock);
+ if (NS_WARN_IF(NS_FAILED(rv))) {
+ DebugOnly<bool> foundInCosts = mCosts.RemoveElementSorted(costEntry);
+ MOZ_ASSERT(foundInCosts, "Lost track of costs for this surface");
+ mTrackingFailureCount++;
+ return false;
+ }
+ }
+
+ mAvailableCost -= costEntry.GetCost();
+ return true;
+ }
+
+ void StopTracking(NotNull<CachedSurface*> aSurface, bool aIsTracked,
+ const StaticMutexAutoLock& aAutoLock) {
+ CostEntry costEntry = aSurface->GetCostEntry();
+
+ if (aSurface->IsLocked()) {
+ MOZ_ASSERT(mLockedCost >= costEntry.GetCost(), "Costs don't balance");
+ mLockedCost -= costEntry.GetCost();
+ // XXX(seth): It'd be nice to use an O(log n) lookup here. This is O(n).
+ MOZ_ASSERT(!mCosts.Contains(costEntry),
+ "Shouldn't have a cost entry for a locked surface");
+ } else {
+ if (MOZ_LIKELY(aSurface->GetExpirationState()->IsTracked())) {
+ MOZ_ASSERT(aIsTracked, "Expiration-tracking a surface unexpectedly!");
+ mExpirationTracker.RemoveObjectLocked(aSurface, aAutoLock);
+ } else {
+ // Our call to AddObject must have failed in StartTracking; most likely
+ // we're in XPCOM shutdown right now.
+ MOZ_ASSERT(!aIsTracked, "Not expiration-tracking an unlocked surface!");
+ }
+
+ DebugOnly<bool> foundInCosts = mCosts.RemoveElementSorted(costEntry);
+ MOZ_ASSERT(foundInCosts, "Lost track of costs for this surface");
+ }
+
+ mAvailableCost += costEntry.GetCost();
+ MOZ_ASSERT(mAvailableCost <= mMaxCost,
+ "More available cost than we started with");
+ }
+
+ LookupResult Lookup(const ImageKey aImageKey, const SurfaceKey& aSurfaceKey,
+ const StaticMutexAutoLock& aAutoLock, bool aMarkUsed) {
+ RefPtr<ImageSurfaceCache> cache = GetImageCache(aImageKey);
+ if (!cache) {
+ // No cached surfaces for this image.
+ return LookupResult(MatchType::NOT_FOUND);
+ }
+
+ RefPtr<CachedSurface> surface = cache->Lookup(aSurfaceKey, aMarkUsed);
+ if (!surface) {
+ // Lookup in the per-image cache missed.
+ return LookupResult(MatchType::NOT_FOUND);
+ }
+
+ if (surface->IsPlaceholder()) {
+ return LookupResult(MatchType::PENDING);
+ }
+
+ DrawableSurface drawableSurface = surface->GetDrawableSurface();
+ if (!drawableSurface) {
+ // The surface was released by the operating system. Remove the cache
+ // entry as well.
+ Remove(WrapNotNull(surface), /* aStopTracking */ true, aAutoLock);
+ return LookupResult(MatchType::NOT_FOUND);
+ }
+
+ if (aMarkUsed &&
+ !MarkUsed(WrapNotNull(surface), WrapNotNull(cache), aAutoLock)) {
+ Remove(WrapNotNull(surface), /* aStopTracking */ false, aAutoLock);
+ return LookupResult(MatchType::NOT_FOUND);
+ }
+
+ MOZ_ASSERT(surface->GetSurfaceKey() == aSurfaceKey,
+ "Lookup() not returning an exact match?");
+ return LookupResult(std::move(drawableSurface), MatchType::EXACT);
+ }
+
+ LookupResult LookupBestMatch(const ImageKey aImageKey,
+ const SurfaceKey& aSurfaceKey,
+ const StaticMutexAutoLock& aAutoLock,
+ bool aMarkUsed) {
+ RefPtr<ImageSurfaceCache> cache = GetImageCache(aImageKey);
+ if (!cache) {
+ // No cached surfaces for this image.
+ return LookupResult(
+ MatchType::NOT_FOUND,
+ SurfaceCache::ClampSize(aImageKey, aSurfaceKey.Size()));
+ }
+
+ // Repeatedly look up the best match, trying again if the resulting surface
+ // has been freed by the operating system, until we can either lock a
+ // surface for drawing or there are no matching surfaces left.
+ // XXX(seth): This is O(N^2), but N is expected to be very small. If we
+ // encounter a performance problem here we can revisit this.
+
+ RefPtr<CachedSurface> surface;
+ DrawableSurface drawableSurface;
+ MatchType matchType = MatchType::NOT_FOUND;
+ IntSize suggestedSize;
+ while (true) {
+ Tie(surface, matchType, suggestedSize) =
+ cache->LookupBestMatch(aSurfaceKey);
+
+ if (!surface) {
+ return LookupResult(
+ matchType, suggestedSize); // Lookup in the per-image cache missed.
+ }
+
+ drawableSurface = surface->GetDrawableSurface();
+ if (drawableSurface) {
+ break;
+ }
+
+ // The surface was released by the operating system. Remove the cache
+ // entry as well.
+ Remove(WrapNotNull(surface), /* aStopTracking */ true, aAutoLock);
+ }
+
+ MOZ_ASSERT_IF(matchType == MatchType::EXACT,
+ surface->GetSurfaceKey() == aSurfaceKey);
+ MOZ_ASSERT_IF(
+ matchType == MatchType::SUBSTITUTE_BECAUSE_NOT_FOUND ||
+ matchType == MatchType::SUBSTITUTE_BECAUSE_PENDING,
+ surface->GetSurfaceKey().SVGContext() == aSurfaceKey.SVGContext() &&
+ surface->GetSurfaceKey().Playback() == aSurfaceKey.Playback() &&
+ surface->GetSurfaceKey().Flags() == aSurfaceKey.Flags());
+
+ if (matchType == MatchType::EXACT ||
+ matchType == MatchType::SUBSTITUTE_BECAUSE_BEST) {
+ if (aMarkUsed &&
+ !MarkUsed(WrapNotNull(surface), WrapNotNull(cache), aAutoLock)) {
+ Remove(WrapNotNull(surface), /* aStopTracking */ false, aAutoLock);
+ }
+ }
+
+ return LookupResult(std::move(drawableSurface), matchType, suggestedSize);
+ }
+
+ bool CanHold(const Cost aCost) const { return aCost <= mMaxCost; }
+
+ size_t MaximumCapacity() const { return size_t(mMaxCost); }
+
+ void SurfaceAvailable(NotNull<ISurfaceProvider*> aProvider,
+ const StaticMutexAutoLock& aAutoLock) {
+ if (!aProvider->Availability().IsPlaceholder()) {
+ MOZ_ASSERT_UNREACHABLE("Calling SurfaceAvailable on non-placeholder");
+ return;
+ }
+
+ // Reinsert the provider, requesting that Insert() mark it available. This
+ // may or may not succeed, depending on whether some other decoder has
+ // beaten us to the punch and inserted a non-placeholder version of this
+ // surface first, but it's fine either way.
+ // XXX(seth): This could be implemented more efficiently; we should be able
+ // to just update our data structures without reinserting.
+ Insert(aProvider, /* aSetAvailable = */ true, aAutoLock);
+ }
+
+ void LockImage(const ImageKey aImageKey) {
+ RefPtr<ImageSurfaceCache> cache = GetImageCache(aImageKey);
+ if (!cache) {
+ cache = new ImageSurfaceCache(aImageKey);
+ mImageCaches.Put(aImageKey, RefPtr{cache});
+ }
+
+ cache->SetLocked(true);
+
+ // We don't relock this image's existing surfaces right away; instead, the
+ // image should arrange for Lookup() to touch them if they are still useful.
+ }
+
+ void UnlockImage(const ImageKey aImageKey,
+ const StaticMutexAutoLock& aAutoLock) {
+ RefPtr<ImageSurfaceCache> cache = GetImageCache(aImageKey);
+ if (!cache || !cache->IsLocked()) {
+ return; // Already unlocked.
+ }
+
+ cache->SetLocked(false);
+ DoUnlockSurfaces(WrapNotNull(cache), /* aStaticOnly = */ false, aAutoLock);
+ }
+
+ void UnlockEntries(const ImageKey aImageKey,
+ const StaticMutexAutoLock& aAutoLock) {
+ RefPtr<ImageSurfaceCache> cache = GetImageCache(aImageKey);
+ if (!cache || !cache->IsLocked()) {
+ return; // Already unlocked.
+ }
+
+ // (Note that we *don't* unlock the per-image cache here; that's the
+ // difference between this and UnlockImage.)
+ DoUnlockSurfaces(WrapNotNull(cache),
+ /* aStaticOnly = */
+ !StaticPrefs::image_mem_animated_discardable_AtStartup(),
+ aAutoLock);
+ }
+
+ already_AddRefed<ImageSurfaceCache> RemoveImage(
+ const ImageKey aImageKey, const StaticMutexAutoLock& aAutoLock) {
+ RefPtr<ImageSurfaceCache> cache = GetImageCache(aImageKey);
+ if (!cache) {
+ return nullptr; // No cached surfaces for this image, so nothing to do.
+ }
+
+ // Discard all of the cached surfaces for this image.
+ // XXX(seth): This is O(n^2) since for each item in the cache we are
+ // removing an element from the costs array. Since n is expected to be
+ // small, performance should be good, but if usage patterns change we should
+ // change the data structure used for mCosts.
+ for (auto iter = cache->ConstIter(); !iter.Done(); iter.Next()) {
+ StopTracking(WrapNotNull(iter.UserData()),
+ /* aIsTracked */ true, aAutoLock);
+ }
+
+ // The per-image cache isn't needed anymore, so remove it as well.
+ // This implicitly unlocks the image if it was locked.
+ mImageCaches.Remove(aImageKey);
+
+ // Since we did not actually remove any of the surfaces from the cache
+ // itself, only stopped tracking them, we should free it outside the lock.
+ return cache.forget();
+ }
+
+ void PruneImage(const ImageKey aImageKey,
+ const StaticMutexAutoLock& aAutoLock) {
+ RefPtr<ImageSurfaceCache> cache = GetImageCache(aImageKey);
+ if (!cache) {
+ return; // No cached surfaces for this image, so nothing to do.
+ }
+
+ cache->Prune([this, &aAutoLock](NotNull<CachedSurface*> aSurface) -> void {
+ StopTracking(aSurface, /* aIsTracked */ true, aAutoLock);
+ // Individual surfaces must be freed outside the lock.
+ mCachedSurfacesDiscard.AppendElement(aSurface);
+ });
+
+ MaybeRemoveEmptyCache(aImageKey, cache);
+ }
+
+ void DiscardAll(const StaticMutexAutoLock& aAutoLock) {
+ // Remove in order of cost because mCosts is an array and the other data
+ // structures are all hash tables. Note that locked surfaces are not
+ // removed, since they aren't present in mCosts.
+ while (!mCosts.IsEmpty()) {
+ Remove(mCosts.LastElement().Surface(), /* aStopTracking */ true,
+ aAutoLock);
+ }
+ }
+
+ void DiscardForMemoryPressure(const StaticMutexAutoLock& aAutoLock) {
+ // Compute our discardable cost. Since locked surfaces aren't discardable,
+ // we exclude them.
+ const Cost discardableCost = (mMaxCost - mAvailableCost) - mLockedCost;
+ MOZ_ASSERT(discardableCost <= mMaxCost, "Discardable cost doesn't add up");
+
+ // Our target is to raise our available cost by (1 / mDiscardFactor) of our
+ // discardable cost - in other words, we want to end up with about
+ // (discardableCost / mDiscardFactor) fewer bytes stored in the surface
+ // cache after we're done.
+ const Cost targetCost = mAvailableCost + (discardableCost / mDiscardFactor);
+
+ if (targetCost > mMaxCost - mLockedCost) {
+ MOZ_ASSERT_UNREACHABLE("Target cost is more than we can discard");
+ DiscardAll(aAutoLock);
+ return;
+ }
+
+ // Discard surfaces until we've reduced our cost to our target cost.
+ while (mAvailableCost < targetCost) {
+ MOZ_ASSERT(!mCosts.IsEmpty(), "Removed everything and still not done");
+ Remove(mCosts.LastElement().Surface(), /* aStopTracking */ true,
+ aAutoLock);
+ }
+ }
+
+ void TakeDiscard(nsTArray<RefPtr<CachedSurface>>& aDiscard,
+ const StaticMutexAutoLock& aAutoLock) {
+ MOZ_ASSERT(aDiscard.IsEmpty());
+ aDiscard = std::move(mCachedSurfacesDiscard);
+ }
+
+ void LockSurface(NotNull<CachedSurface*> aSurface,
+ const StaticMutexAutoLock& aAutoLock) {
+ if (aSurface->IsPlaceholder() || aSurface->IsLocked()) {
+ return;
+ }
+
+ StopTracking(aSurface, /* aIsTracked */ true, aAutoLock);
+
+ // Lock the surface. This can fail.
+ aSurface->SetLocked(true);
+ DebugOnly<bool> tracked = StartTracking(aSurface, aAutoLock);
+ MOZ_ASSERT(tracked);
+ }
+
+ size_t ShallowSizeOfIncludingThis(
+ MallocSizeOf aMallocSizeOf, const StaticMutexAutoLock& aAutoLock) const {
+ size_t bytes =
+ aMallocSizeOf(this) + mCosts.ShallowSizeOfExcludingThis(aMallocSizeOf) +
+ mImageCaches.ShallowSizeOfExcludingThis(aMallocSizeOf) +
+ mCachedSurfacesDiscard.ShallowSizeOfExcludingThis(aMallocSizeOf) +
+ mExpirationTracker.ShallowSizeOfExcludingThis(aMallocSizeOf);
+ for (auto iter = mImageCaches.ConstIter(); !iter.Done(); iter.Next()) {
+ bytes += iter.UserData()->ShallowSizeOfIncludingThis(aMallocSizeOf);
+ }
+ return bytes;
+ }
+
+ NS_IMETHOD
+ CollectReports(nsIHandleReportCallback* aHandleReport, nsISupports* aData,
+ bool aAnonymize) override {
+ StaticMutexAutoLock lock(sInstanceMutex);
+
+ uint32_t lockedImageCount = 0;
+ uint32_t totalSurfaceCount = 0;
+ uint32_t lockedSurfaceCount = 0;
+ for (auto iter = mImageCaches.ConstIter(); !iter.Done(); iter.Next()) {
+ totalSurfaceCount += iter.UserData()->Count();
+ if (iter.UserData()->IsLocked()) {
+ ++lockedImageCount;
+ }
+ for (auto surfIter = iter.UserData()->ConstIter(); !surfIter.Done();
+ surfIter.Next()) {
+ if (surfIter.UserData()->IsLocked()) {
+ ++lockedSurfaceCount;
+ }
+ }
+ }
+
+ // clang-format off
+ // We have explicit memory reporting for the surface cache which is more
+ // accurate than the cost metrics we report here, but these metrics are
+ // still useful to report, since they control the cache's behavior.
+ MOZ_COLLECT_REPORT(
+ "explicit/images/cache/overhead", KIND_HEAP, UNITS_BYTES,
+ ShallowSizeOfIncludingThis(SurfaceCacheMallocSizeOf, lock),
+"Memory used by the surface cache data structures, excluding surface data.");
+
+ MOZ_COLLECT_REPORT(
+ "imagelib-surface-cache-estimated-total",
+ KIND_OTHER, UNITS_BYTES, (mMaxCost - mAvailableCost),
+"Estimated total memory used by the imagelib surface cache.");
+
+ MOZ_COLLECT_REPORT(
+ "imagelib-surface-cache-estimated-locked",
+ KIND_OTHER, UNITS_BYTES, mLockedCost,
+"Estimated memory used by locked surfaces in the imagelib surface cache.");
+
+ MOZ_COLLECT_REPORT(
+ "imagelib-surface-cache-tracked-cost-count",
+ KIND_OTHER, UNITS_COUNT, mCosts.Length(),
+"Total number of surfaces tracked for cost (and expiry) in the imagelib surface cache.");
+
+ MOZ_COLLECT_REPORT(
+ "imagelib-surface-cache-tracked-expiry-count",
+ KIND_OTHER, UNITS_COUNT, mExpirationTracker.Length(lock),
+"Total number of surfaces tracked for expiry (and cost) in the imagelib surface cache.");
+
+ MOZ_COLLECT_REPORT(
+ "imagelib-surface-cache-image-count",
+ KIND_OTHER, UNITS_COUNT, mImageCaches.Count(),
+"Total number of images in the imagelib surface cache.");
+
+ MOZ_COLLECT_REPORT(
+ "imagelib-surface-cache-locked-image-count",
+ KIND_OTHER, UNITS_COUNT, lockedImageCount,
+"Total number of locked images in the imagelib surface cache.");
+
+ MOZ_COLLECT_REPORT(
+ "imagelib-surface-cache-image-surface-count",
+ KIND_OTHER, UNITS_COUNT, totalSurfaceCount,
+"Total number of surfaces in the imagelib surface cache.");
+
+ MOZ_COLLECT_REPORT(
+ "imagelib-surface-cache-locked-surfaces-count",
+ KIND_OTHER, UNITS_COUNT, lockedSurfaceCount,
+"Total number of locked surfaces in the imagelib surface cache.");
+
+ MOZ_COLLECT_REPORT(
+ "imagelib-surface-cache-overflow-count",
+ KIND_OTHER, UNITS_COUNT, mOverflowCount,
+"Count of how many times the surface cache has hit its capacity and been "
+"unable to insert a new surface.");
+
+ MOZ_COLLECT_REPORT(
+ "imagelib-surface-cache-tracking-failure-count",
+ KIND_OTHER, UNITS_COUNT, mTrackingFailureCount,
+"Count of how many times the surface cache has failed to begin tracking a "
+"given surface.");
+
+ MOZ_COLLECT_REPORT(
+ "imagelib-surface-cache-already-present-count",
+ KIND_OTHER, UNITS_COUNT, mAlreadyPresentCount,
+"Count of how many times the surface cache has failed to insert a surface "
+"because it is already present.");
+
+ MOZ_COLLECT_REPORT(
+ "imagelib-surface-cache-table-failure-count",
+ KIND_OTHER, UNITS_COUNT, mTableFailureCount,
+"Count of how many times the surface cache has failed to insert a surface "
+"because a hash table could not accept an entry.");
+ // clang-format on
+
+ return NS_OK;
+ }
+
+ void CollectSizeOfSurfaces(const ImageKey aImageKey,
+ nsTArray<SurfaceMemoryCounter>& aCounters,
+ MallocSizeOf aMallocSizeOf,
+ const StaticMutexAutoLock& aAutoLock) {
+ RefPtr<ImageSurfaceCache> cache = GetImageCache(aImageKey);
+ if (!cache) {
+ return; // No surfaces for this image.
+ }
+
+ // Report all surfaces in the per-image cache.
+ cache->CollectSizeOfSurfaces(
+ aCounters, aMallocSizeOf,
+ [this, &aAutoLock](NotNull<CachedSurface*> aSurface) -> void {
+ StopTracking(aSurface, /* aIsTracked */ true, aAutoLock);
+ // Individual surfaces must be freed outside the lock.
+ mCachedSurfacesDiscard.AppendElement(aSurface);
+ });
+
+ MaybeRemoveEmptyCache(aImageKey, cache);
+ }
+
+ void ReleaseImageOnMainThread(already_AddRefed<image::Image>&& aImage,
+ const StaticMutexAutoLock& aAutoLock) {
+ RefPtr<image::Image> image = aImage;
+ if (!image) {
+ return;
+ }
+
+ bool needsDispatch = mReleasingImagesOnMainThread.IsEmpty();
+ mReleasingImagesOnMainThread.AppendElement(image);
+
+ if (!needsDispatch) {
+ // There is already a ongoing task for ClearReleasingImages().
+ return;
+ }
+
+ NS_DispatchToMainThread(NS_NewRunnableFunction(
+ "SurfaceCacheImpl::ReleaseImageOnMainThread",
+ []() -> void { SurfaceCache::ClearReleasingImages(); }));
+ }
+
+ void TakeReleasingImages(nsTArray<RefPtr<image::Image>>& aImage,
+ const StaticMutexAutoLock& aAutoLock) {
+ MOZ_ASSERT(NS_IsMainThread());
+ aImage.SwapElements(mReleasingImagesOnMainThread);
+ }
+
+ private:
+ already_AddRefed<ImageSurfaceCache> GetImageCache(const ImageKey aImageKey) {
+ RefPtr<ImageSurfaceCache> imageCache;
+ mImageCaches.Get(aImageKey, getter_AddRefs(imageCache));
+ return imageCache.forget();
+ }
+
+ void MaybeRemoveEmptyCache(const ImageKey aImageKey,
+ ImageSurfaceCache* aCache) {
+ // Remove the per-image cache if it's unneeded now. Keep it if the image is
+ // locked, since the per-image cache is where we store that state. Note that
+ // we don't push it into mImageCachesDiscard because all of its surfaces
+ // have been removed, so it is safe to free while holding the lock.
+ if (aCache->IsEmpty() && !aCache->IsLocked()) {
+ mImageCaches.Remove(aImageKey);
+ }
+ }
+
+ // This is similar to CanHold() except that it takes into account the costs of
+ // locked surfaces. It's used internally in Insert(), but it's not exposed
+ // publicly because we permit multithreaded access to the surface cache, which
+ // means that the result would be meaningless: another thread could insert a
+ // surface or lock an image at any time.
+ bool CanHoldAfterDiscarding(const Cost aCost) const {
+ return aCost <= mMaxCost - mLockedCost;
+ }
+
+ bool MarkUsed(NotNull<CachedSurface*> aSurface,
+ NotNull<ImageSurfaceCache*> aCache,
+ const StaticMutexAutoLock& aAutoLock) {
+ if (aCache->IsLocked()) {
+ LockSurface(aSurface, aAutoLock);
+ return true;
+ }
+
+ nsresult rv = mExpirationTracker.MarkUsedLocked(aSurface, aAutoLock);
+ if (NS_WARN_IF(NS_FAILED(rv))) {
+ // If mark used fails, it is because it failed to reinsert the surface
+ // after removing it from the tracker. Thus we need to update our
+ // own accounting but otherwise expect it to be untracked.
+ StopTracking(aSurface, /* aIsTracked */ false, aAutoLock);
+ return false;
+ }
+ return true;
+ }
+
+ void DoUnlockSurfaces(NotNull<ImageSurfaceCache*> aCache, bool aStaticOnly,
+ const StaticMutexAutoLock& aAutoLock) {
+ AutoTArray<NotNull<CachedSurface*>, 8> discard;
+
+ // Unlock all the surfaces the per-image cache is holding.
+ for (auto iter = aCache->ConstIter(); !iter.Done(); iter.Next()) {
+ NotNull<CachedSurface*> surface = WrapNotNull(iter.UserData());
+ if (surface->IsPlaceholder() || !surface->IsLocked()) {
+ continue;
+ }
+ if (aStaticOnly &&
+ surface->GetSurfaceKey().Playback() != PlaybackType::eStatic) {
+ continue;
+ }
+ StopTracking(surface, /* aIsTracked */ true, aAutoLock);
+ surface->SetLocked(false);
+ if (MOZ_UNLIKELY(!StartTracking(surface, aAutoLock))) {
+ discard.AppendElement(surface);
+ }
+ }
+
+ // Discard any that we failed to track.
+ for (auto iter = discard.begin(); iter != discard.end(); ++iter) {
+ Remove(*iter, /* aStopTracking */ false, aAutoLock);
+ }
+ }
+
+ void RemoveEntry(const ImageKey aImageKey, const SurfaceKey& aSurfaceKey,
+ const StaticMutexAutoLock& aAutoLock) {
+ RefPtr<ImageSurfaceCache> cache = GetImageCache(aImageKey);
+ if (!cache) {
+ return; // No cached surfaces for this image.
+ }
+
+ RefPtr<CachedSurface> surface =
+ cache->Lookup(aSurfaceKey, /* aForAccess = */ false);
+ if (!surface) {
+ return; // Lookup in the per-image cache missed.
+ }
+
+ Remove(WrapNotNull(surface), /* aStopTracking */ true, aAutoLock);
+ }
+
+ class SurfaceTracker final
+ : public ExpirationTrackerImpl<CachedSurface, 2, StaticMutex,
+ StaticMutexAutoLock> {
+ public:
+ explicit SurfaceTracker(uint32_t aSurfaceCacheExpirationTimeMS)
+ : ExpirationTrackerImpl<CachedSurface, 2, StaticMutex,
+ StaticMutexAutoLock>(
+ aSurfaceCacheExpirationTimeMS, "SurfaceTracker") {}
+
+ protected:
+ void NotifyExpiredLocked(CachedSurface* aSurface,
+ const StaticMutexAutoLock& aAutoLock) override {
+ sInstance->Remove(WrapNotNull(aSurface), /* aStopTracking */ true,
+ aAutoLock);
+ }
+
+ void NotifyHandlerEndLocked(const StaticMutexAutoLock& aAutoLock) override {
+ sInstance->TakeDiscard(mDiscard, aAutoLock);
+ }
+
+ void NotifyHandlerEnd() override {
+ nsTArray<RefPtr<CachedSurface>> discard(std::move(mDiscard));
+ }
+
+ StaticMutex& GetMutex() override { return sInstanceMutex; }
+
+ nsTArray<RefPtr<CachedSurface>> mDiscard;
+ };
+
+ class MemoryPressureObserver final : public nsIObserver {
+ public:
+ NS_DECL_ISUPPORTS
+
+ NS_IMETHOD Observe(nsISupports*, const char* aTopic,
+ const char16_t*) override {
+ nsTArray<RefPtr<CachedSurface>> discard;
+ {
+ StaticMutexAutoLock lock(sInstanceMutex);
+ if (sInstance && strcmp(aTopic, "memory-pressure") == 0) {
+ sInstance->DiscardForMemoryPressure(lock);
+ sInstance->TakeDiscard(discard, lock);
+ }
+ }
+ return NS_OK;
+ }
+
+ private:
+ virtual ~MemoryPressureObserver() {}
+ };
+
+ nsTArray<CostEntry> mCosts;
+ nsRefPtrHashtable<nsPtrHashKey<Image>, ImageSurfaceCache> mImageCaches;
+ nsTArray<RefPtr<CachedSurface>> mCachedSurfacesDiscard;
+ SurfaceTracker mExpirationTracker;
+ RefPtr<MemoryPressureObserver> mMemoryPressureObserver;
+ nsTArray<RefPtr<image::Image>> mReleasingImagesOnMainThread;
+ const uint32_t mDiscardFactor;
+ const Cost mMaxCost;
+ Cost mAvailableCost;
+ Cost mLockedCost;
+ size_t mOverflowCount;
+ size_t mAlreadyPresentCount;
+ size_t mTableFailureCount;
+ size_t mTrackingFailureCount;
+};
+
+NS_IMPL_ISUPPORTS(SurfaceCacheImpl, nsIMemoryReporter)
+NS_IMPL_ISUPPORTS(SurfaceCacheImpl::MemoryPressureObserver, nsIObserver)
+
+///////////////////////////////////////////////////////////////////////////////
+// Public API
+///////////////////////////////////////////////////////////////////////////////
+
+/* static */
+void SurfaceCache::Initialize() {
+ // Initialize preferences.
+ MOZ_ASSERT(NS_IsMainThread());
+ MOZ_ASSERT(!sInstance, "Shouldn't initialize more than once");
+
+ // See StaticPrefs for the default values of these preferences.
+
+ // Length of time before an unused surface is removed from the cache, in
+ // milliseconds.
+ uint32_t surfaceCacheExpirationTimeMS =
+ StaticPrefs::image_mem_surfacecache_min_expiration_ms_AtStartup();
+
+ // What fraction of the memory used by the surface cache we should discard
+ // when we get a memory pressure notification. This value is interpreted as
+ // 1/N, so 1 means to discard everything, 2 means to discard about half of the
+ // memory we're using, and so forth. We clamp it to avoid division by zero.
+ uint32_t surfaceCacheDiscardFactor =
+ max(StaticPrefs::image_mem_surfacecache_discard_factor_AtStartup(), 1u);
+
+ // Maximum size of the surface cache, in kilobytes.
+ uint64_t surfaceCacheMaxSizeKB =
+ StaticPrefs::image_mem_surfacecache_max_size_kb_AtStartup();
+
+ if (sizeof(uintptr_t) <= 4) {
+ // Limit surface cache to 1 GB if our address space is 32 bit.
+ surfaceCacheMaxSizeKB = 1024 * 1024;
+ }
+
+ // A knob determining the actual size of the surface cache. Currently the
+ // cache is (size of main memory) / (surface cache size factor) KB
+ // or (surface cache max size) KB, whichever is smaller. The formula
+ // may change in the future, though.
+ // For example, a value of 4 would yield a 256MB cache on a 1GB machine.
+ // The smallest machines we are likely to run this code on have 256MB
+ // of memory, which would yield a 64MB cache on this setting.
+ // We clamp this value to avoid division by zero.
+ uint32_t surfaceCacheSizeFactor =
+ max(StaticPrefs::image_mem_surfacecache_size_factor_AtStartup(), 1u);
+
+ // Compute the size of the surface cache.
+ uint64_t memorySize = PR_GetPhysicalMemorySize();
+ if (memorySize == 0) {
+ MOZ_ASSERT_UNREACHABLE("PR_GetPhysicalMemorySize not implemented here");
+ memorySize = 256 * 1024 * 1024; // Fall back to 256MB.
+ }
+ uint64_t proposedSize = memorySize / surfaceCacheSizeFactor;
+ uint64_t surfaceCacheSizeBytes =
+ min(proposedSize, surfaceCacheMaxSizeKB * 1024);
+ uint32_t finalSurfaceCacheSizeBytes =
+ min(surfaceCacheSizeBytes, uint64_t(UINT32_MAX));
+
+ // Create the surface cache singleton with the requested settings. Note that
+ // the size is a limit that the cache may not grow beyond, but we do not
+ // actually allocate any storage for surfaces at this time.
+ sInstance = new SurfaceCacheImpl(surfaceCacheExpirationTimeMS,
+ surfaceCacheDiscardFactor,
+ finalSurfaceCacheSizeBytes);
+ sInstance->InitMemoryReporter();
+}
+
+/* static */
+void SurfaceCache::Shutdown() {
+ RefPtr<SurfaceCacheImpl> cache;
+ {
+ StaticMutexAutoLock lock(sInstanceMutex);
+ MOZ_ASSERT(NS_IsMainThread());
+ MOZ_ASSERT(sInstance, "No singleton - was Shutdown() called twice?");
+ cache = sInstance.forget();
+ }
+}
+
+/* static */
+LookupResult SurfaceCache::Lookup(const ImageKey aImageKey,
+ const SurfaceKey& aSurfaceKey,
+ bool aMarkUsed) {
+ nsTArray<RefPtr<CachedSurface>> discard;
+ LookupResult rv(MatchType::NOT_FOUND);
+
+ {
+ StaticMutexAutoLock lock(sInstanceMutex);
+ if (!sInstance) {
+ return rv;
+ }
+
+ rv = sInstance->Lookup(aImageKey, aSurfaceKey, lock, aMarkUsed);
+ sInstance->TakeDiscard(discard, lock);
+ }
+
+ return rv;
+}
+
+/* static */
+LookupResult SurfaceCache::LookupBestMatch(const ImageKey aImageKey,
+ const SurfaceKey& aSurfaceKey,
+ bool aMarkUsed) {
+ nsTArray<RefPtr<CachedSurface>> discard;
+ LookupResult rv(MatchType::NOT_FOUND);
+
+ {
+ StaticMutexAutoLock lock(sInstanceMutex);
+ if (!sInstance) {
+ return rv;
+ }
+
+ rv = sInstance->LookupBestMatch(aImageKey, aSurfaceKey, lock, aMarkUsed);
+ sInstance->TakeDiscard(discard, lock);
+ }
+
+ return rv;
+}
+
+/* static */
+InsertOutcome SurfaceCache::Insert(NotNull<ISurfaceProvider*> aProvider) {
+ nsTArray<RefPtr<CachedSurface>> discard;
+ InsertOutcome rv(InsertOutcome::FAILURE);
+
+ {
+ StaticMutexAutoLock lock(sInstanceMutex);
+ if (!sInstance) {
+ return rv;
+ }
+
+ rv = sInstance->Insert(aProvider, /* aSetAvailable = */ false, lock);
+ sInstance->TakeDiscard(discard, lock);
+ }
+
+ return rv;
+}
+
+/* static */
+bool SurfaceCache::CanHold(const IntSize& aSize,
+ uint32_t aBytesPerPixel /* = 4 */) {
+ StaticMutexAutoLock lock(sInstanceMutex);
+ if (!sInstance) {
+ return false;
+ }
+
+ Cost cost = ComputeCost(aSize, aBytesPerPixel);
+ return sInstance->CanHold(cost);
+}
+
+/* static */
+bool SurfaceCache::CanHold(size_t aSize) {
+ StaticMutexAutoLock lock(sInstanceMutex);
+ if (!sInstance) {
+ return false;
+ }
+
+ return sInstance->CanHold(aSize);
+}
+
+/* static */
+void SurfaceCache::SurfaceAvailable(NotNull<ISurfaceProvider*> aProvider) {
+ StaticMutexAutoLock lock(sInstanceMutex);
+ if (!sInstance) {
+ return;
+ }
+
+ sInstance->SurfaceAvailable(aProvider, lock);
+}
+
+/* static */
+void SurfaceCache::LockImage(const ImageKey aImageKey) {
+ StaticMutexAutoLock lock(sInstanceMutex);
+ if (sInstance) {
+ return sInstance->LockImage(aImageKey);
+ }
+}
+
+/* static */
+void SurfaceCache::UnlockImage(const ImageKey aImageKey) {
+ StaticMutexAutoLock lock(sInstanceMutex);
+ if (sInstance) {
+ return sInstance->UnlockImage(aImageKey, lock);
+ }
+}
+
+/* static */
+void SurfaceCache::UnlockEntries(const ImageKey aImageKey) {
+ StaticMutexAutoLock lock(sInstanceMutex);
+ if (sInstance) {
+ return sInstance->UnlockEntries(aImageKey, lock);
+ }
+}
+
+/* static */
+void SurfaceCache::RemoveImage(const ImageKey aImageKey) {
+ RefPtr<ImageSurfaceCache> discard;
+ {
+ StaticMutexAutoLock lock(sInstanceMutex);
+ if (sInstance) {
+ discard = sInstance->RemoveImage(aImageKey, lock);
+ }
+ }
+}
+
+/* static */
+void SurfaceCache::PruneImage(const ImageKey aImageKey) {
+ nsTArray<RefPtr<CachedSurface>> discard;
+ {
+ StaticMutexAutoLock lock(sInstanceMutex);
+ if (sInstance) {
+ sInstance->PruneImage(aImageKey, lock);
+ sInstance->TakeDiscard(discard, lock);
+ }
+ }
+}
+
+/* static */
+void SurfaceCache::DiscardAll() {
+ nsTArray<RefPtr<CachedSurface>> discard;
+ {
+ StaticMutexAutoLock lock(sInstanceMutex);
+ if (sInstance) {
+ sInstance->DiscardAll(lock);
+ sInstance->TakeDiscard(discard, lock);
+ }
+ }
+}
+
+/* static */
+void SurfaceCache::CollectSizeOfSurfaces(
+ const ImageKey aImageKey, nsTArray<SurfaceMemoryCounter>& aCounters,
+ MallocSizeOf aMallocSizeOf) {
+ nsTArray<RefPtr<CachedSurface>> discard;
+ {
+ StaticMutexAutoLock lock(sInstanceMutex);
+ if (!sInstance) {
+ return;
+ }
+
+ sInstance->CollectSizeOfSurfaces(aImageKey, aCounters, aMallocSizeOf, lock);
+ sInstance->TakeDiscard(discard, lock);
+ }
+}
+
+/* static */
+size_t SurfaceCache::MaximumCapacity() {
+ StaticMutexAutoLock lock(sInstanceMutex);
+ if (!sInstance) {
+ return 0;
+ }
+
+ return sInstance->MaximumCapacity();
+}
+
+/* static */
+bool SurfaceCache::IsLegalSize(const IntSize& aSize) {
+ // reject over-wide or over-tall images
+ const int32_t k64KLimit = 0x0000FFFF;
+ if (MOZ_UNLIKELY(aSize.width > k64KLimit || aSize.height > k64KLimit)) {
+ NS_WARNING("image too big");
+ return false;
+ }
+
+ // protect against invalid sizes
+ if (MOZ_UNLIKELY(aSize.height <= 0 || aSize.width <= 0)) {
+ return false;
+ }
+
+ // check to make sure we don't overflow a 32-bit
+ CheckedInt32 requiredBytes =
+ CheckedInt32(aSize.width) * CheckedInt32(aSize.height) * 4;
+ if (MOZ_UNLIKELY(!requiredBytes.isValid())) {
+ NS_WARNING("width or height too large");
+ return false;
+ }
+ return true;
+}
+
+IntSize SurfaceCache::ClampVectorSize(const IntSize& aSize) {
+ // If we exceed the maximum, we need to scale the size downwards to fit.
+ // It shouldn't get here if it is significantly larger because
+ // VectorImage::UseSurfaceCacheForSize should prevent us from requesting
+ // a rasterized version of a surface greater than 4x the maximum.
+ int32_t maxSizeKB =
+ StaticPrefs::image_cache_max_rasterized_svg_threshold_kb();
+ if (maxSizeKB <= 0) {
+ return aSize;
+ }
+
+ int64_t proposedKB = int64_t(aSize.width) * aSize.height / 256;
+ if (maxSizeKB >= proposedKB) {
+ return aSize;
+ }
+
+ double scale = sqrt(double(maxSizeKB) / proposedKB);
+ return IntSize(int32_t(scale * aSize.width), int32_t(scale * aSize.height));
+}
+
+IntSize SurfaceCache::ClampSize(ImageKey aImageKey, const IntSize& aSize) {
+ if (aImageKey->GetType() != imgIContainer::TYPE_VECTOR) {
+ return aSize;
+ }
+
+ return ClampVectorSize(aSize);
+}
+
+/* static */
+void SurfaceCache::ReleaseImageOnMainThread(
+ already_AddRefed<image::Image> aImage, bool aAlwaysProxy) {
+ if (NS_IsMainThread() && !aAlwaysProxy) {
+ RefPtr<image::Image> image = std::move(aImage);
+ return;
+ }
+
+ StaticMutexAutoLock lock(sInstanceMutex);
+ if (sInstance) {
+ sInstance->ReleaseImageOnMainThread(std::move(aImage), lock);
+ } else {
+ NS_ReleaseOnMainThread("SurfaceCache::ReleaseImageOnMainThread",
+ std::move(aImage), /* aAlwaysProxy */ true);
+ }
+}
+
+/* static */
+void SurfaceCache::ClearReleasingImages() {
+ MOZ_ASSERT(NS_IsMainThread());
+
+ nsTArray<RefPtr<image::Image>> images;
+ {
+ StaticMutexAutoLock lock(sInstanceMutex);
+ if (sInstance) {
+ sInstance->TakeReleasingImages(images, lock);
+ }
+ }
+}
+
+} // namespace image
+} // namespace mozilla